a16z: Global AI Has Sparked an Explosion in Apps, but Users and Paying Customers Haven't Kept Up...

AI is making software supply cheap, yet users and revenue are not growing alongside it. As "building the product" becomes ever easier, the truly scarce capabilities have shifted to the opposite end: uncovering demand, capturing attention, getting people to pay, and staying alive when capital becomes expensive.

A collection of 13 charts compiled by a16z captures this dislocation. App stores are flooded with more new releases, yet most still fail to clear the cold-start hurdle; AI productivity tools have absorbed the most striking growth, but paying out of pocket has not yet become a mainstream consumer habit; a handful of young companies are rushing to unicorn status faster, while companies that scaled up during the previous expansion wave are entering a concentrated shakeout. Data from software equities and Massachusetts further extends this dilemma to mature enterprises and real-world operating environments: an influx of technical supply does not automatically generate commercial demand of an equivalent scale.

Code Is No Longer Scarce; User Attention Has Become the True Bottleneck

Research cited by a16z labels the period after February 2025 as the "agentic coding era," when AI began shouldering more continuous programming tasks. Since then, new monthly app additions on iOS climbed from approximately 30,000–50,000 to roughly 120,000, while Chrome extensions surged from about 3,000 to over 20,000. Android also rebounded from its trough to roughly 100,000, though it still has not surpassed its all-time high around 2021; the "roughly 2–4x" mentioned in the original source is better suited to describe changes relative to recent lows across iOS, Chrome, and Android, rather than implying that all three platforms set new long-term records.

The commonality lies on the demand side: while new app releases exploded, total user ratings and downloads failed to increase in tandem. The barrier to building products has fallen, but finding that first cohort of real users has only become more congested.

a16z uses "escape velocity" to describe a new application clearing the most fundamental cold-start thresholds: at least 10 ratings on iOS, over 100 downloads on Android, and at least 10 downloads for Chrome extensions. Relative to January 2025, the share of new releases reaching these thresholds dropped by approximately 8, 3, and 14 percentage points, respectively. AI can help developers generate and ship products faster, but it does not solve distribution or demand on their behalf.

Nor has the broader app market expanded noticeably as a result of increased supply. Sensor Tower's tracking across all app categories in the U.S. shows that from December 2024 to August 2026, total revenue grew by approximately 2%, while total time spent increased by 7%. As more apps flood the stores, they are still competing for slices of a consumer pie whose growth remains modest.

Breaking down the aggregate market, the value of AI is already evident, yet highly concentrated. Time spent in productivity tools grew by roughly 55% and revenue surged by approximately 110%, primarily driven by ChatGPT, Claude, Gemini, and Grok. Developer tools also expanded rapidly, though their overall market size remains small, with Replit standing out as the most prominent product; food & drink revenue grew by roughly 125%, but from a similarly small base, and time spent remained virtually flat.

At the other end of the market, contraction is underway: gaming time spent declined by approximately 3% and revenue fell by about 15%, lifestyle revenue dropped by roughly 30%, and books revenue decreased by about 40%. The new revenue generated by a handful of top AI applications has been partially offset by declines across these mature, heavyweight categories. Rather than a scenario where "all AI applications prosper together," the market exhibits growth heavily concentrated among a select few winners.

What truly holds back consumer AI is that mainstream willingness to pay has yet to materialize. In the first quarter of 2023, the proportion paying out of pocket for AI across all age groups was below or near 1%; by the first quarter of 2026, the overall level reported by Consumer Edge had climbed to roughly 3%. Growth has been distinctly concentrated among younger cohorts: approximately 6% for 18-to-24-year-olds versus about 1.4% for those aged 65 and older—a gap of nearly 4x.

This chart tracks U.S. consumers purchasing AI services using personal credit or debit cards, excluding corporate or employer procurement. It indicates that direct individual payment remains in an early stage, while also demonstrating that consumer AI is making its earliest inroads onto the paid subscriptions list of younger generations.

A Handful of New Companies Move Faster, While More Tech Firms Trade Growth for Survival

While the consumer side is still developing paying habits, the venture capital ecosystem has witnessed a cohort of new companies accelerating their growth. Data from SVB reveals that the median age of all active unicorns rose from roughly 12 years to nearly 15 years, whereas the median age of new unicorns minted that year upon reaching a $1 billion valuation fell from nearly 7 years to about 4.3 years—a 37% shortening compared to 2023. Existing companies are aging as a whole, while new entrants are clearing the valuation threshold much sooner.

Valuation is not the only thing moving faster. Excluding mega-scale AI compute companies, businesses founded in 2022 saw their median revenue rise from $2.8 million to $5.6 million between Year 3 and Year 4—a net gain of approximately $2.8 million; earlier cohorts typically added only about $1.0 million over the same timeframe. Measured by incremental revenue, the ramp-up velocity of this newer cohort is nearly triple that of the past.

However, disparities within the same cohort are also substantial: the middle 50% of the 2022 vintage in Year 4 spanned roughly from $500,000 to $13 million. A rapidly climbing median does not mean the entire cohort is taking off in unison; rather, it suggests that this new generation of companies has already developed a wide divergence between winners and laggards.

Whether this growth is sustainable is the real question going forward. The original authors propose one possibility: the 2022 cohort may contain many "AI wrappers" that scaled quickly but have yet to prove their moats. Whether the 2023 and 2024 vintages can replicate this trajectory will determine whether this represents a structural advantage for a new generation of companies or merely an early windfall.

While a handful of new companies are accelerating, the broader tech landscape is shifting into defense. Median year-over-year revenue growth rates across fintech, enterprise tech, frontier tech, and consumer internet pulled back from roughly 50%–80% in early 2022 to about 15%–30%; at the same time, the scale of losses has narrowed. By the second quarter of 2026, the median net profit margin stood at approximately -5% for consumer internet, around -7% for fintech, roughly -13% for enterprise tech, and remained around -30% for frontier tech.

Companies that secured funding over the past three quarters present a contrasting profile: their median revenue growth rate remains around 63%–73%, yet net profit margins hover between -43% and -60%. While the broader universe of companies trades slower growth for an operating posture closer to breakeven, the recently funded sample continues to exchange heavy burn for breakneck growth.

The Previous Startup Boom Enters a Shakeout, with B2B SaaS Comprising Over a Quarter of Shutdowns

Companies founded en masse during the previous era of low interest rates are undergoing a concentrated test of survival. Records of failures and shutdowns among U.S. VC-backed companies climbed from 893 in 2023 to 1,602 in 2024, and 1,981 in 2025. Dashed lines are used for 2025 and 2026 in the chart, with 2026 marked at 2,345; however, because the full annual estimation methodology is not detailed, this figure is better understood as an indicative trend value.

Among the 2026 shutdown sample with known founding years, companies founded between 2019 and 2021 accounted for 34%, representing the single largest group; those founded between 2022 and 2023 made up 20%, and those founded after 2024 represented 9%. The core of this wave of closures remains older cohorts left over from previous peak fundraising years.

Sector breakdowns reveal that the targets of this shakeout are also evolving. In SimpleClosure's shutdown sample, the share of AI startups declined from 17.7% in 2024 to 15.9% in 2025, and fell further to 14.4% in the first half of 2026; meanwhile, B2B SaaS accounted for 27.3% of the shutdown sample in the first half of 2026, standing as the largest distinct category.

B2B SaaS was originally among the most startup-dense tracks, meaning industry inventory naturally influences the composition of the shutdown sample. The more direct takeaway is that while AI companies account for a diminishing share of shutdowns, traditional B2B software represents over a quarter—demonstrating that a large volume of previous-generation software startups are still bearing the brunt of ongoing liquidations.

Software Stocks Have Rebounded, but Operating Pressure Surfaces First in Next-Quarter Guidance

In February 2026, Citrini published an article predicting that AI would bring an end to the software industry. Indexed to 100 in mid-February, stock indices for large-, mid-, and small-cap software companies continued to slide, bottoming out between roughly 91 and 94 in April; by mid-September, small caps had rebounded to about 130, mid caps to about 134, and large caps to about 127. The pessimistic article coincided closely with the tail end of that downturn, and software stocks have since achieved a pronounced recovery.

Following the price rebound, the next question is whether the pessimistic expectations have manifested in actual business performance. Across the first two quarters of 2026, approximately 94% of software companies in the tracked sample met or beat market expectations for actual performance; current-period results have not suffered widespread misses.

The specific breakdown is as follows (percentages may not sum to 100% due to rounding):

• Q1 Actual Results | 146 companies: 66% beat, 28% met, 6% missed.
• Q2 Actual Results | 148 companies: 66% beat, 28% met, 7% missed.
• Q3 Forward Guidance | 69 companies: 30% above, 49% in line, 20% below.

Pressure is appearing first in the subsequent quarter: the proportion of Q3 guidance falling below expectations climbed to 20%, while the beat rate dropped from 66% in the prior two quarters to 30%. Although the sample size narrowed from around 150 companies to 69, the trend is evident—while current results largely held the line against market expectations, management teams are sending distinctly more cautious signals for the quarter ahead.

Massachusetts Faces High Startup Costs Alongside a Recent Net Decline in Business Establishments

The final set of charts grounds the issue in specific operating environments. Total fees for establishing a Limited Liability Company (LLC) vary widely across states: Massachusetts stands at roughly $1,000, five times the national median of around $200; California is $890, Nevada $775, Tennessee $600, Delaware $510, and Montana is the lowest at just $35.

A thousand dollars will not single-handedly dictate whether a company stays or leaves; the original authors view it as a signal of broader business burdens, citing CNBC's Cost of Doing Business rankings, where Massachusetts ranks 49th nationwide. For small businesses, combined costs including registration, labor, and taxes jointly shape decisions about where to incorporate and expand.

Shifts in local business establishments have coincided with these high costs. Data from the BLS Business Employment Dynamics shows that the birth-to-death ratio of private establishments in Massachusetts has weakened over the long term, though the decline has not been linear: it once surged to a ten-year peak of approximately 1.60 in 2021, before falling to roughly 0.75 in 2023 and 2024. A ratio of 0.75 means that for every 3 new establishments created, roughly 4 disappear.

According to calculations by the Pioneer Institute based on BLS data, Massachusetts recorded a net loss of approximately 17,500 establishments across 9 consecutive quarters. While this figure does not represent the "number of companies relocating out," it clearly demonstrates that the local pace of replenishing new establishments is lagging behind the rate of attrition. LLC fees are only one component of the business climate; together, these two data points illustrate that cost pressures and net establishment declines coexist, rather than suggesting that a single fee drove the entire shift.

Collectively, these 13 charts point to a single shift: the bottleneck in the software industry during the AI era is pivoting away from producing code and toward capturing demand, driving willingness to pay, maintaining capital efficiency, and weathering operating costs. The ability to build more applications is no longer scarce; retaining users, getting them to pay, and sustaining company operations under financial and cost pressures are the capabilities that are truly difficult to replicate.